Long-term Human Capability Preservation in Agentic Coding Tools

Determine how to design and integrate mechanisms within agentic coding systems such as Claude Code that explicitly support long-term human improvement, deeper understanding, and sustained codebase coherence, beyond short-term capability amplification.

Background

The paper positions Claude Code as a powerful agentic coding tool that substantially amplifies users’ short-term capabilities. However, it notes a gap in explicit architectural mechanisms that help users maintain or grow long-term skills, comprehension, and codebase coherence. This concern becomes a cross-cutting evaluative lens applied across the architecture and motivates several future directions related to sustainability of human capability.

By flagging this gap as an explicit open question early in the paper, the authors set the stage for later sections that discuss observability, governance, memory, and horizon scaling as related axes where design answers may help address the long-term sustainability problem.

References

Used as an evaluative lens, our study also reveals an open question: while the Claude Code agent system substantially amplifies the short-term capabilities of programmers and end users, it offers limited mechanisms that explicitly support long-term human improvement, deeper understanding, and sustained codebase coherence.

— Dive into Claude Code: The Design Space of Today's and Future AI Agent Systems  (2604.14228 - Liu et al., 14 Apr 2026) in Section 1 (Introduction)

Note that nothing in the CAT paradigm requires the outer agent to remain the smaller model; it is an empirical question whether the demands of scientific judgement must match or exceed those of engineering execution in the long run.

— Training AI Scientists to Replicate Research  (2608.13331 - Falck et al., 13 Aug 2026) in Discussion, paragraph “Coding agent as a tool (CAT)”